WO2025089036A1 - Method for producing property determination model, property determination model, property determination method, and property determination device - Google Patents

Abstract

Provided is a method for producing a property determination model. In the method, when a property determination model with which a property of a subject is determined is produced on the basis of expression levels of a plurality of small RNA molecules, a specimen from which the plurality of small RNA molecules is derived is selected. A specimen, in which the conditions until the measurement of the expression levels of the plurality of small RNA molecules in the specimen after the collection of the specimen from the subject satisfy predetermined conditions, is selected as a reference subject. Using the expression levels of the plurality of small RNA molecules in the reference specimen, the property determination model is produced.

Description

Property determination model generation method, property determination model, property determination method, and property determination device

This disclosure relates to a method for generating a property determination model, a property determination model, a property determination method, and a property determination device.

A method is known in which sample data is obtained that includes the expression levels of multiple types of microRNA in blood samples measured by a next-generation sequencer (hereinafter referred to as NGS), and the presence or absence of a disease is determined using a trained model that has been machine-learned using the microRNA expression level data of disease-free samples and multiple disease samples as training data (see Non-Patent Document 1).

Non-patent document 1: Suzuki, K., Igata, H., Abe, M., Yamamoto, Y., small RNA based cancer classification project, Multiple cancer type classific ation by small RNA expression profiles with plasma samples from multiple facilities. Cancer Sci. 113, 2144-2166 (26 February 2022).

Here, small RNAs such as microRNAs are present in trace amounts in samples and are also very unstable. Therefore, if the conditions between when a sample is collected from a subject and when the expression levels of multiple small RNAs in the sample are measured are inappropriate, the measured value of the small RNA expression level may not reflect the expression level of the small RNAs contained in the sample at the time of collection.

In addition, even if the conditions from collection of a sample from a subject to measurement of the expression levels of multiple small RNAs in the sample are appropriate, if various conditions differ between the multiple samples, the values expressed as small RNA data indicating the measured values for each sample may differ between samples, even if the measured values reflect the expression levels of the small RNAs contained in the sample at the time of collection. Examples of such various conditions include the measurement device itself that measures the expression levels of multiple small RNAs, the measurement conditions of the measurement device, and the processing conditions in the preprocessing performed before measuring the expression levels of multiple small RNAs (e.g., the reagents used in the preprocessing).

Therefore, if a property determination model for determining a subject's property (e.g., disease) is generated using the expression levels of small RNAs in samples measured under inappropriate conditions, or the expression levels of small RNAs measured under different conditions between samples, the property determination model will be generated based on information (i.e., the expression levels of small RNAs) that does not reflect the small RNAs in the sample at the time of collection, i.e., the subject, and the property determination model generated may have low determination accuracy.

The objective of this disclosure is to make it possible to generate a property determination model with high determination accuracy.

A method for generating a property determination model according to one embodiment of the present disclosure is a method for generating a property determination model for determining the property of a subject based on the expression levels of multiple small RNAs, in which a sample from which the multiple small RNAs are derived is selected, and a sample that satisfies predetermined conditions from the time the sample is collected from the subject to the time the expression levels of the multiple small RNAs in the sample are measured is selected as a standard sample, and the property determination model is generated using the expression levels of the multiple small RNAs in the standard sample.

According to this disclosure, it is possible to generate a property determination model with high determination accuracy.

FIG. 1 is a flow chart showing each step of a method for generating a disease trained model according to this embodiment. FIG. 2 is a flow chart showing each step of the disease determination method according to the present embodiment. FIG. 2 is a schematic block diagram of an example of a computer that functions as the disease determination device of this embodiment. 1 is a block diagram showing a functional configuration of an illness determination device according to an embodiment of the present invention; 11 is a result table showing the determination results in this embodiment.

Below, an example of an embodiment of the technology of the present disclosure will be described with reference to the drawings. Note that components and processes that perform the same operation, action, and function will be given the same reference numerals throughout the drawings, and duplicated explanations may be omitted as appropriate. Each drawing is merely a schematic illustration to allow a sufficient understanding of the technology of the present disclosure. Therefore, the technology of the present disclosure is not limited to the illustrated examples. Furthermore, in this embodiment, explanations of configurations that are not directly related to the present disclosure or well-known configurations may be omitted.

<Disease trained model generation method 10>
First, a disease trained model generation method 10 according to the present embodiment will be described. Fig. 1 is a schematic diagram showing each step of the disease trained model generation method 10 according to the present embodiment. The disease trained model generation method 10 is an example of a property determination model generation method.

The disease-trained model generation method 10 is a generation method for selecting a specimen from which multiple small RNAs are derived when generating a disease-trained model for determining a disease in a subject based on the expression levels of multiple small RNAs, and selects a specimen that satisfies predetermined conditions from the time the specimen is collected from the subject until the expression levels of multiple small RNAs in the specimen are measured as a standard specimen, and generates a disease-trained model using the expression levels of multiple small RNAs in the standard specimen.

Note that "selection" is not limited to selecting some samples from multiple samples, but also includes selecting all samples from multiple samples.

A standard sample is a sample that serves as a reference for generating a disease-trained model, and is a sample in which the conditions from when the sample is collected from a subject to when the expression levels of multiple small RNAs in the sample are measured satisfy predetermined conditions. Specifically, a standard sample is a sample in which at least one of the collection conditions for collecting a sample from a subject, the storage conditions for storing the collected sample, and the measurement conditions for measuring the expression levels of multiple small RNAs in the subject satisfy predetermined conditions. Specific examples of the collection conditions include various conditions in the various processes in the collection step 11 described below. Specific examples of the storage conditions include various conditions in the various processes in the storage step 13 described below. Specific examples of the measurement conditions include various conditions in the various processes in the measurement step 12 described below.

Examples of specimens collected from subjects include body fluids, cells, extracellular vesicles, and tissue fragments. Examples of body fluids include blood, serum, urine, tears, saliva, sweat, semen, lymph, tissue fluid, body cavity fluid (e.g., pleural fluid, ascites, etc.), cerebrospinal fluid, amniotic fluid, vaginal fluid, and nasal mucus. Examples of cells include red blood cells, white blood cells, and platelets. Examples of extracellular vesicles include exosomes and liposomes. Examples of tissue fragments include FFPE (Formalin Fixed Paraffin Embedded) specimens, biopsy specimens, and frozen specimens. Note that specimens may be specimens from which multiple small RNAs are derived, in other words, specimens from which the expression levels of multiple small RNAs can be measured. The term "subject" refers to a concept that indicates the subject from whom a specimen is collected. The subject from whom a specimen is collected may be a human or a non-human animal. Non-human animals include non-human mammals (monkeys, dogs, cats, mice, rats, rabbits, cows, horses, pigs, sheep, etc.), birds (chickens, quails, etc.), etc.

An example of a small RNA is microRNA. Note that the small RNA may also be a small RNA other than microRNA (e.g., piRNA and tsRNA).

In this embodiment, the disease trained model generation method 10 is a method for generating a disease trained model used in the disease assessment method 100 described below, which determines the presence or absence of a disease, by inputting small RNA data indicating the results of measuring the expression levels of multiple small RNAs in a sample from a subject into the disease trained model. Note that the generation method 10 is a generation method for generating a disease trained model as an example of a property assessment model for determining the property of a subject, but the property of the subject to be assessed is not limited to the subject's disease. In other words, the property may be any property that can be assessed based on the expression levels of multiple small RNAs. Examples of property assessment include confirming the efficacy of the subject's drug, determining the possibility of disease recurrence, confirming/assessing lifestyle habits such as checking the presence or absence of a smoking history or drinking history, and predicting physical age.

Specifically, as shown in FIG. 1, the disease trained model generation method 10 includes a collection step 11, a storage step 13, a measurement step 12, a selection step 14, and a model generation step 17. In this embodiment, the collection step 11, the storage step 13, the measurement step 12, the selection step 14, and the model generation step 17 are performed in this order, for example. Note that in this embodiment, the disease trained model generation method 10 is an example of a property determination model generation method, but the selection step 14 and the model generation step 17, which are part of the disease trained model generation method 10, may be understood as an example of a property determination model generation method.

Below, we will explain each step of the disease trained model generation method 10, the disease prevalence determination method 100 using the disease trained model, and the property determination system 20 that executes the disease prevalence determination method 100.

<Collecting step 11>
The collecting step 11 is a step of collecting a body fluid as a specimen from a subject. The collecting step 11 is performed before the execution of the storage step 13 and the measurement step 12. In the case where serum is collected as the body fluid in the collecting step 11, for example, the collection process is performed as follows.

First, blood is collected using a vacuum blood collection tube containing a serum separating agent (blood collection process).
Next, immediately after the blood collection process, the blood is mixed by inversion.
The blood is then allowed to coagulate at room temperature for at least 30 minutes.
Next, the mixture is centrifuged and the serum is separated (centrifugation treatment).
Then, the serum is separated and stored at −80° C. (storage treatment).

In the collection process 11, the process from collecting bodily fluids from the subject to completing the centrifugation process (specifically, the process from the blood collection process to the preservation process described above) is performed within a predetermined time (e.g., 2 hours). Note that the centrifugation process is, for example, the process from the centrifugation process to the preservation process described above.

Here, the longer the processing time from collection of bodily fluids from the subject to completion of the centrifugation operation (hereinafter, sometimes referred to as collection processing time), the more the bodily fluids deteriorate, and the more the measurement results in the measurement step 12 change. Specifically, as the collection processing time becomes longer, small RNAs such as microRNAs contained in the bodily fluids are degraded, and the expression level of small RNAs measured in the measurement step 12 may decrease. In addition, as the collection processing time becomes longer, small RNAs leak from formed components (e.g., blood cells and platelets) that settle by centrifugation in the bodily fluids, and the amount of small RNAs present in the bodily fluids increases. Specifically, when the bodily fluid is serum, if the time to coagulate blood or the processing time from centrifugation to separation of serum becomes longer, small RNAs leak from formed components, and the expression level of small RNAs measured in the measurement step 12 may increase. Furthermore, as the collection processing time becomes longer, small RNAs are degraded, and the length of the base sequence to be measured falls below the threshold and becomes undetectable. In this way, the degradation of small RNA may change the amount of small RNA present measured in the measurement process 12. This phenomenon becomes noticeable when the collection process time exceeds a predetermined time, and affects the measurement results of the small RNA measured in the measurement process 12. For this reason, in the collection process 11, it is required that the collection process be performed within the predetermined time (e.g., 2 hours), for example.

The collection process 11 may include at least a process of collecting bodily fluid from the subject and a process of centrifuging the bodily fluid.

<Preservation step 13>
The preservation step 13 is a step of preserving the collected body fluid at a predetermined preservation temperature. In this embodiment, the collected body fluid (e.g., serum) in the collection step 11 is preserved at a predetermined preservation temperature (e.g., −80° C.) to prevent the body fluid from deteriorating before the measurement step 12 is performed. Note that the preservation temperature is not limited to −80° C., and can be set as appropriate, for example, at a temperature lower than 4° C. In addition, in the preservation step 13, the body fluid collected from the subject may be transported, for example, to a measurement location where the measurement step 12 is performed, while being preserved at a predetermined preservation temperature.

Here, if the storage temperature is high and the storage time is long, the body fluid (e.g., serum) may deteriorate, causing a change in the measurement result in the measurement step 12. Specifically, if the storage temperature is high and the storage time is long, small RNAs such as microRNAs contained in the body fluids are easily decomposed, and the expression level of the small RNA measured in the measurement step 12 may decrease. Also, if the storage temperature is high and the storage time is long, the small RNA is decomposed and the length of the measured base sequence falls below the threshold and becomes undetectable. In this way, the amount of small RNA present measured in the measurement step 12 may change due to the decomposition of small RNA. This phenomenon is noticeable when the storage temperature and storage time exceed predetermined values, and affects the measurement result of the small RNA measured in the measurement step 12. For this reason, in the storage step 13, it is required that the storage temperature and storage time are within predetermined values to store the body fluid.

The storage temperatures can be classified into a first temperature range (e.g., a temperature range below 4°C) in which small RNAs undergo little change regardless of storage time, a second temperature range (e.g., a temperature range between 4°C and 25°C) in which small RNAs undergo change depending on storage time, and a third temperature range (e.g., a temperature range above 25°C) in which small RNAs undergo change regardless of storage time.

<Measurement step 12>
The measurement step 12 is a step of measuring the expression levels of multiple small RNAs in the subject's body fluid. Specifically, in the measurement step 12, a next-generation sequencer (NGS) is used as a measurement device to measure multiple small RNAs contained in the subject's body fluid and identify the base sequence of each small RNA. Next, the number of identified small RNAs is counted for each base sequence to obtain the number of reads of the small RNA in the NGS. This number of reads of the small RNA corresponds to the expression level (specifically, the absolute expression level) of the small RNA. In other words, this number of reads of the small RNA means small RNA data showing the result of measuring the expression levels of multiple small RNAs in the subject's body fluid. However, although the expression levels of multiple small RNAs (e.g., microRNAs) in body fluids (e.g., blood) are absolute values derived from a living organism, it is difficult to quantify the expression levels of small RNAs in body fluids as absolute values because it is necessary to quantify the expression levels through a process of a measurement device or reagent treatment. Therefore, the expression levels of small RNAs may be expressed as relative values by processing the measurement results from NGS, and such relative values also refer to small RNA data.

Furthermore, in order to measure the absolute expression level, the number of reads of small RNA output by NGS may be normalized to obtain a relative expression level (i.e., relative expression level). As a means of normalization, for example, RPM (Read Per Million) normalization or normalization using a small RNA as an internal standard may be used. In this way, the small RNA data may indicate a normalized relative expression level. Note that the small RNA data may be an absolute quantitative value quantified as an absolute value.

In addition, NGS makes it possible to simultaneously measure the expression levels of multiple small RNAs in multiple body fluids (e.g., blood samples taken from multiple subjects). That is, in measurement step 12, the expression levels of multiple small RNAs are measured for multiple body fluids in the same process.

In addition to next-generation sequencers, other measuring devices such as DNA chips, quantitative PCR, and flow cytometers can also be used as long as the expression levels of multiple small RNAs can be measured. As described above, if the expression levels of multiple small RNAs can be measured and a small RNA showing the measurement results can be obtained, various methods, including publicly known methods, can be used to measure the expression levels of multiple small RNAs.

The measurement step 12 may include a pretreatment step performed before measuring the expression levels of the multiple small RNAs in the subject's body fluid. The pretreatment step may include various pretreatment steps performed on the sample (e.g., centrifugation, storage, library preparation, etc.). The various pretreatment steps may be performed using various reagents.

<Sorting step 14>
The selection process 14 is a process of selecting a sample as a standard sample, the conditions from when the sample is collected from the subject to when the expression levels of multiple small RNAs in the sample are measured (hereinafter referred to as selection conditions) satisfying predetermined conditions.

<First selection condition>
The selection condition is the processing time from collection of a body fluid as a sample from a subject to completion of centrifugation (hereinafter, sometimes referred to as collection processing time). That is, in the selection step 14, from among the multiple samples in which the expression levels of multiple small RNAs were measured in the measurement step 12, a sample with an appropriate collection processing time is selected as a standard sample.

Since small RNAs are present in trace amounts in samples and are very unstable, if the collection and processing time is inappropriate, the measurement results for the small RNA expression level may not accurately reflect the expression level of the small RNA contained in the sample.

Specifically, as mentioned above, when the collection and processing time exceeds a predetermined time (e.g., 2 hours), the phenomenon that the expression level of the measured small RNA changes due to deterioration of the sample, etc., becomes evident. Therefore, in the selection process 14, as an example, samples whose collection and processing time is within the predetermined time (e.g., 2 hours) are selected as standard samples.

The selection of standard samples is performed, for example, by the creator who creates the disease-learned model (i.e., the executor who executes this creation method 10) managing the collection processing time and selecting samples whose collection processing time is within the predetermined time (e.g., 2 hours). In this way, the selection of standard samples can be performed by a method that depends on the creator, in which the creator manages the collection processing time, as one example, but is not limited to this method. For example, the selection of standard samples may be performed using a collection processing time-learned model.

Specifically, the small RNA data is input into a collection processing time trained model that has been machine-learned to learn the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and collection processing time data showing the collection processing time of the body fluids collected from the multiple subjects, thereby examining the appropriateness of the collection processing time, and selecting body fluids for which the collection processing time is appropriate as standard samples. The small RNA data is data showing the measurement results of the expression levels of small RNAs measured in the measurement process 12.

The measurement data is specifically the result of measuring multiple small RNAs in bodily fluids collected from multiple subjects, including subjects with a disease and subjects without a disease (i.e., healthy subjects). The measurement data can be obtained by measuring using a measurement method similar to that of the measurement step 12 described above. The collection processing time data is specifically data indicating the suitability of the collection processing time from when the bodily fluid related to the measurement data is collected from the subject to when the centrifugation operation is completed.

Specifically, as the measurement data to be machine-learned (hereinafter, sometimes referred to as training data), for example, the first measurement data obtained by performing a collection process on a body fluid collected from a healthy subject within a predetermined time (e.g., 2 hours), the second measurement data obtained by performing a collection process on a body fluid collected from a healthy subject for more than a predetermined time (e.g., 2 hours), and the third measurement data obtained by performing a collection process on a body fluid collected from a subject suffering from a disease within a predetermined time (e.g., 2 hours) are used to generate a processing time trained model. Note that as the training data to be machine-learned, the collection processing time trained model may be generated using four measurement data, which are the above-mentioned first measurement data, second measurement data, and third measurement data, plus the measurement data obtained by performing a collection process on a body fluid collected from a subject suffering from a disease for more than a predetermined time (e.g., 2 hours). In this way, in this embodiment, not only the measurement data of the body fluid of a healthy subject but also the measurement data of the body fluid of a subject suffering from a disease are used as training data.

<Second selection condition>
Further, the selection conditions are the storage temperature and storage time at the storage temperature of the body fluid as a sample collected from the subject. That is, in the selection step 14, a sample having an appropriate storage temperature and storage time is selected as a standard sample from among the multiple samples in which the expression levels of multiple small RNAs were measured in the measurement step 12. In this manner, multiple conditions may be set as the selection conditions.

As with the collection and processing time described above, if the storage temperature of the sample collected from the subject and the storage time at that storage temperature are inappropriate, the measurement results of the small RNA expression level may not accurately reflect the expression level of the small RNA contained in the sample.

Specifically, as mentioned above, when the storage temperature and storage time exceed predetermined values, the phenomenon in which the expression level of the measured small RNA changes due to deterioration of the sample, etc., becomes evident. Therefore, in the selection process 14, samples whose storage temperature and storage time are below the predetermined values are selected as standard samples.

Considering the storage temperature, there are three possible temperature ranges: a first temperature range (e.g., a temperature range below 4°C) in which small RNAs undergo little change regardless of storage time; a second temperature range (e.g., a temperature range between 4°C and 25°C) in which small RNAs undergo change depending on storage time; and a third temperature range (e.g., a temperature range above 25°C) in which small RNAs undergo change regardless of storage time.

Specifically, in the selection process 14, a sample stored at a temperature in the first temperature range (e.g., -80°C) is selected as the standard sample. In this case, the storage time is not an issue.

The selection of standard specimens is performed, for example, by the creator of the disease-trained model managing the storage temperature and storage time, and selecting specimens whose storage temperature and storage time are equal to or less than the predetermined values. In this way, the selection of standard specimens can be performed by a method that depends on the creator, in which the creator manages the storage temperature and storage time, as one example, but is not limited to this method. For example, the selection of standard specimens may be performed using a storage condition-trained model.

Specifically, the small RNA data is input into a storage condition trained model that has been machine-learned to learn the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and storage condition data showing the storage temperature and storage time at that storage temperature of the body fluids collected from the multiple subjects, and the small RNA data is input into the storage condition trained model to check the suitability of the storage temperature and storage time at that storage temperature, and body fluids with appropriate storage temperature and storage time are selected as standard samples. The small RNA data is data showing the measurement results of the expression levels of small RNAs measured in the measurement process 12.

The measurement data is specifically the result of measuring multiple small RNAs in samples collected from multiple subjects, including subjects with a disease and subjects without a disease (i.e., healthy subjects). The measurement data can be obtained by a measurement method similar to that of the measurement step 12 described above. The storage condition data is specifically data indicating the suitability of the storage conditions, taking into account both the storage temperature of the sample related to the measurement data and the storage time at that storage temperature.

Specifically, the storage condition learned model is generated using the measurement data (hereinafter sometimes referred to as training data) to be machine-learned, which are the first measurement data obtained by storing the body fluid collected from a healthy subject under appropriate storage conditions, the second measurement data obtained by storing the body fluid collected from a healthy subject under inappropriate storage conditions, and the third measurement data obtained by storing the body fluid collected from a subject suffering from a disease under appropriate storage conditions. In this manner, in this embodiment, not only the measurement data of the healthy subject's sample but also the measurement data of the subject suffering from a disease is used as the training data. Note that the storage condition learned model may be generated using four measurement data as the training data to be machine-learned, which are the above-mentioned first measurement data, second measurement data, and third measurement data, plus the measurement data obtained by storing the body fluid collected from a subject suffering from a disease under inappropriate storage conditions. Note that the above-mentioned appropriate storage conditions are, for example, when stored at a temperature belonging to the first temperature range (for example, -80°C), and the above-mentioned inappropriate storage conditions are, for example, when stored at a temperature higher than the temperature belonging to the first temperature range.

<Third Selection Condition>
Furthermore, in the case where blood is collected from a subject as a sample, the selection condition is the degree of hemolysis of the blood sample collected from the subject. That is, in the selection step 14, a sample having an appropriate degree of hemolysis is selected as a standard sample from among the multiple samples in which the expression levels of multiple small RNAs are measured in the measurement step 12.

As with the collection processing time, storage temperature, and storage time described above, if the degree of hemolysis of the blood sample collected from the subject is inappropriate, the measurement results of the small RNA expression level may not accurately reflect the expression level of the small RNA contained in the sample.

Specifically, when the degree of hemolysis exceeds a predetermined value, the phenomenon in which the expression level of the measured small RNA changes becomes evident. Therefore, in this production method, samples with a degree of hemolysis below a predetermined value are selected as standard samples.

The selection of standard specimens is performed, for example, by the creator of the disease-trained model managing the degree of hemolysis and selecting specimens whose degree of hemolysis is equal to or lower than the predetermined value. In this way, the selection of standard specimens can be performed, for example, by a method that depends on the creator, in which the creator manages the degree of hemolysis, but is not limited to this method. For example, the selection of standard specimens may be performed using a model trained on the degree of hemolysis.

Specifically, the small RNA data is input into a hemolysis level trained model that has been machine-learned to learn the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in blood collected from multiple subjects and hemolysis level data showing the degree of hemolysis in the blood collected from the multiple subjects, and the small RNA data is input to inspect the degree of hemolysis and select samples with an appropriate degree of hemolysis as standard samples. The small RNA data is data showing the measurement results of the expression levels of small RNAs measured in the measurement process 12.

The measurement data is specifically the result of measuring multiple small RNAs in blood collected from multiple subjects. The multiple subjects may include subjects with a disease and subjects without a disease (i.e., healthy subjects), or may include only subjects with a disease and subjects without a disease (i.e., healthy subjects). The measurement data can be obtained by measurement using a measurement method similar to that of the measurement step 12 described above.

The hemolysis degree data is specifically data indicating whether the degree of hemolysis of the blood related to the measurement data is appropriate. The degree of hemolysis in the hemolysis degree data is based on whether the degree of hemolysis is appropriate for judgment in the disease-learned model used in the disease judgment method 100. In other words, the hemolysis degree data is data that defines a degree of hemolysis at which a correct judgment can be performed in the disease judgment method 100 as appropriate (good), and defines a degree of hemolysis at which a correct judgment cannot be performed in the disease judgment method 100 as inappropriate (bad).

Specifically, as the measurement data to be subjected to machine learning (hereinafter sometimes referred to as training data), a trained model for the degree of hemolysis is generated using measurement data obtained from blood samples taken from multiple subjects under conditions where the degree of hemolysis is appropriate, and measurement data obtained from blood samples taken from multiple subjects under conditions where the degree of hemolysis is inappropriate.

<Other selection criteria>
Furthermore, as selection conditions, the type and lot of blood collection tube, conditions from blood collection to storage such as centrifugation rotation speed, storage conditions, type and lot of small RNA measurement reagent, measurement pretreatment conditions such as reaction temperature, type of measurement device, and measurement conditions such as sample concentration at the time of measurement may be set.

As described above, the selection conditions are the conditions from when a sample is collected from a subject to when the expression levels of multiple small RNAs in the sample are measured. Here, the measurement includes obtaining measurement data that indicates the measurement results of the expression levels and is used to generate a disease-learned model. Therefore, if the measurement data is processed, for example, before the generation of a disease-learned model, the processing can also be set as a selection condition.

In addition, in the selection step 14, as described above, the sample whose selection conditions satisfy the predetermined conditions is selected as the standard sample, and the predetermined conditions may be that the conditions from when the sample is collected from the subject to when the expression levels of the multiple small RNAs in the sample are measured are constant among the multiple samples. In other words, in the selection step 14, the multiple samples whose selection conditions are constant among the multiple samples may be selected as the standard samples. Specifically, for example, multiple samples in which the expression levels of multiple small RNAs have been measured using the same measurement device under the same measurement conditions can be selected as the standard samples.

The selection conditions are set so that the measurement data used to generate the disease-trained model can be acquired as data reflecting the expression levels of small RNA contained in the sample at the time of collection, and the values expressed as data indicating the expression levels are constant across multiple samples. In other words, the selection conditions can be said to be conditions that make it possible to acquire the aforementioned data as the measurement data.

<Model generation step 17>
The model generation step 17 is a step of generating a disease-trained model using the expression levels of multiple small RNAs in the standard specimen selected in the selection step 14. In one example, the model generation step 17 generates a disease-trained model by machine learning the correlation between measurement data indicating the results of measuring the expression levels of multiple small RNAs in a standard specimen and morbidity data indicating the presence or absence of a disease in the subject from whom the standard specimen was collected.

Specifically, a disease-learned model is generated using measurement data obtained from standard specimens of subjects with a disease and measurement data obtained from standard specimens of subjects without a disease (i.e., healthy individuals) as the measurement data to be subjected to machine learning (hereinafter sometimes referred to as training data).

In addition, in the model generation step 17, measurement data for the top 100 small RNAs in expression levels in standard samples from multiple subjects is used as training data.

In this way, in this embodiment, measurement data is used that excludes small RNAs with relatively low expression levels. Small RNAs with relatively low expression levels have the advantage that they can easily function as explanatory variables, since even small fluctuations in expression level result in large fold changes. However, they also have the disadvantage of being easily affected by fluctuations due to factors other than the element of interest (specifically, the collection and processing time) and fluctuations due to measurement errors, making them less robust. Therefore, by using only small RNAs with relatively high expression levels as training data, the effect of strengthening robustness can be achieved.

In addition, in the model generation step 17, for example, measurement data excluding small RNAs whose expression level was zero in a standard specimen from any of the subjects is used as training data. Since the variation from zero has a large impact as the fold change is infinite, small RNAs whose expression level is zero in any of the specimens used as training data are excluded from the training data.

As described above, in this embodiment, a disease trained model is generated by having a learning model learn the presence or absence of a disease qualitatively. The disease trained model then determines the presence or absence of a disease, and outputs the presence or absence, thereby judging the presence or absence of a disease. Therefore, in the disease judgment method 100 described below, the test result indicating the presence or absence of a disease is displayed.

In this embodiment, the disease-trained model is generated by having the learning model learn the presence or absence of a disease qualitatively, but this is not limited to the above. For example, the disease-trained model may be generated by having the learning model learn the subject of judgment quantitatively. In this case, the disease-trained model judges, for example, the probability of having a disease, and the judgment result is output as a judgment value. Then, the presence or absence of a disease is judged based on the judgment result of the judgment unit that judges the presence or absence of a disease based on a comparison between the output judgment value and a threshold value.

<Infection determination method 100>
The disease determination method 100 according to this embodiment will be described. Fig. 2 is a schematic diagram showing each step of the disease determination method 100 according to this embodiment. Note that the disease determination method 100 is an example of a property determination method.

The disease assessment method 100 is a method for assessing the presence or absence of a disease by inputting small RNA data, which indicates the results of measuring the expression levels of multiple small RNAs in a subject's sample, into a disease-trained model.

Specifically, as shown in FIG. 2, the disease determination method 100 has a collection step 111, a storage step 113, a measurement step 112, a selection step 114, and a disease determination step 116. In this embodiment, the collection step 111, the storage step 113, the measurement step 112, the selection step 114, and the disease determination step 116 are performed in this order, for example. Note that in this embodiment, the disease determination method 100 is an example of a property determination method, but the disease determination step 116, which is part of the disease determination method 100, may also be understood as an example of a property determination method.

In the disease assessment method 100, the presence or absence of a disease is assessed as an example of a subject's property, but the property of the subject to be assessed is not limited to the subject's disease. In other words, the property may be any property that can be assessed based on the expression levels of multiple small RNAs. As described above, examples of property assessment include confirming the efficacy of the subject's medication, determining the possibility of disease recurrence, confirming/assessing lifestyle habits such as checking the presence or absence of a smoking history or drinking history, and predicting physical age.

<Collecting step 111, storing step 113, measuring step 112, sorting step 114>
As an example, the collection step 111, the storage step 113, the measurement step 112, and the selection step 114 are performed in the same manner as the collection step 11, the storage step 13, the measurement step 12, and the selection step 14 in the production method 10.

In other words, in the disease assessment method 100, each step from collecting a sample from a subject to measuring the expression levels of multiple small RNAs in the sample (specifically, collection step 111, storage step 113, and measurement step 112) is performed under the same conditions as those in each step in the production method 10 (specifically, collection step 11, storage step 13, and measurement step 12) (in other words, the predetermined conditions in the selection step 14).

Then, in the selection step 114, from among the multiple samples in which the expression levels of multiple small RNAs were measured in the measurement step 112, those samples whose conditions from the collection of the sample from the subject to the measurement of the expression levels of multiple small RNAs in the sample satisfy predetermined conditions (i.e., the same conditions as the predetermined conditions in the selection step 14 of the production method 10) are selected as samples to be assessed in the disease assessment step 116.

<Medition determination step 116>
The disease determination step 116 determines the presence or absence of a disease by inputting multiple small RNA data in the specimen selected in the selection step 114 into a disease trained model.

Therefore, the small RNA data is small RNA data obtained from a sample in which each step from collection of the sample from the subject to measurement of the expression levels of multiple small RNAs in the sample satisfies predetermined conditions (i.e., the same conditions as the predetermined conditions in selection step 14 of production method 10).

The disease trained model is a disease trained model generated by the above-mentioned generation method 10. That is, the disease trained model is generated based on the expression levels of multiple small RNAs, and a sample from which the multiple small RNAs are derived is selected during the generation, and the sample that satisfies predetermined conditions from the time the sample is collected from the subject to the time the expression levels of the multiple small RNAs in the sample are measured is selected as a standard sample, and the disease trained model is generated using the expression levels of the multiple small RNAs in the standard sample. The disease trained model is an example of a property determination model that determines the property of a subject.

In the disease determination process 116, the presence or absence of a disease is determined for the specimen selected in the selection process 114. In other words, the disease determination process 116 is not performed for specimens that were not selected in the selection process 114, i.e., specimens that do not satisfy the selection conditions. In this case, for example, the fact that the selection conditions were not satisfied may be presented to the user (i.e., the person executing the disease determination method 100).

It is desirable that the disease determined by the disease trained model is the same as the disease used for training in the processing time trained model, storage condition trained model, and degree of hemolysis trained model. For example, if the disease used for training in the processing time trained model, storage condition trained model, and degree of hemolysis trained model is cancer, the disease determined by the disease discrimination trained model is also cancer. Specifically, it is desirable that the measurement data used in the disease discrimination trained model is the same as the measurement data used in the processing time trained model, storage condition trained model, degree of hemolysis trained model, and preservative addition time trained model.

As described above, the disease determination step 116 was not performed for samples that were not selected in the selection step 114, but this is not limited to the above. For example, the disease determination step 116 may be performed for samples that were not selected in the selection step 114, for example, if the determination result is to be obtained as reference data. In this case, for example, the user may be informed that a determination was made even though the selection conditions were not satisfied. Also, in this embodiment, the disease determination step 116 was performed after the selection step 114, but it may be performed before the selection step 114. In this case, if the selection conditions are not satisfied in the selection step 114, the determination result of the disease determination step 116 is treated as, for example, reference data.

<Property Determination System 20>
Next, a property determination system 20 will be described as a system for executing the above-mentioned disease determination method 100, which is an example of a property determination method. The property determination system 20 includes a measurement device 21 and a disease determination device 30, as shown in FIG.

<Measuring device 21>
The measurement device 21 is a device that executes the above-mentioned measurement step 112. That is, the measurement device 21 measures the expression levels of a plurality of small RNAs contained in each of a plurality of samples. As the measurement device 21, for example, an NGS is used.

<Medition determination device 30>
The disease determination device 30 executes the above-mentioned selection step 114. That is, the disease determination device 30 selects, from among a plurality of samples in which the expression levels of a plurality of small RNAs have been measured by the measurement device 21, a sample in which the conditions from collection of the sample from the subject to measurement of the expression levels of a plurality of small RNAs in the sample satisfy a predetermined condition as a sample to be determined. The disease determination device 30 is an example of a property determination device. Thus, in this embodiment, the disease determination device 30 is an example of a property determination device, but the property determination system 20 including the measurement device 21 may be understood as an example of a property determination device.

The disease assessment device 30 further executes the disease assessment step 116 described above. That is, the presence or absence of a disease is assessed by inputting small RNA data indicating the expression levels of multiple small RNAs in the selected specimen into the disease-trained model generated by the generation method 10 described above.

The disease assessment device 30 functions as a computer, and as shown in FIG. 3, has a CPU (Central Processing Unit) 31, a ROM (Read Only Memory) 32, a RAM (Random Access Memory) 33, storage 34, an input unit 35, a display unit 36, and a communication interface (I/F) 37. Each component is connected to each other via a bus 39 so that they can communicate with each other.

CPU 31 (an example of a processor) is a central processing unit that executes various programs and controls each part. That is, CPU 31 reads a program from ROM 32 or storage 34, and executes the program using RAM 33 as a working area. CPU 31 controls each of the above components and performs various calculation processes according to the program stored in ROM 32 or storage 34. CPU 31 is an example of a processor.

ROM 32 records various programs and various data. RAM 33 temporarily stores programs or data as a working area. Storage 34 is composed of a HDD (Hard Disk Drive) or SSD (Solid State Drive), and records various programs including the operating system, and various data.

In this embodiment, for example, a disease assessment program for executing a disease assessment process that performs the disease assessment method described above is recorded in storage 34. The disease assessment program may be a single program, or a group of programs consisting of multiple programs or modules. The disease assessment program may be recorded in ROM 32. ROM 32 and storage 34 function as an example of a non-transitory recording medium.

An example of a processor is not limited to the aforementioned CPU, which is a general-purpose processor, but may be, for example, a dedicated processor made up of a circuit designed specifically to execute a specific process. Also, an example of a processor is not limited to a single processor, but may be a processor made up of multiple processors working together at physically separate locations.

The input unit 35 includes a pointing device such as a mouse and a keyboard, and is used to perform various inputs. The input unit 35 also receives as input information on the expression levels of multiple small RNAs measured by the measurement device 21.

The display unit 36 is, for example, a liquid crystal display, and displays various information. In the disease assessment device 30, for example, the selection results and the results of the assessment of the presence or absence of disease can be presented to the user through the display unit 36. The display unit 36 may also function as the input unit 35 by employing a touch panel system.

The communication interface 37 is an interface for communicating with other devices, and uses standards such as Ethernet (registered trademark), FDDI (Fiber Distributed Data Interface), and Wi-Fi (registered trademark).

As shown in FIG. 4, in the disease assessment device 30, the CPU 31 executes the disease assessment program to function as a selection unit 160 and a disease assessment unit 170.

The selection unit 160 executes the selection step 114 described above. That is, the selection unit 160 selects, from among a plurality of samples in which the expression levels of a plurality of small RNAs have been measured by the measurement device 21, a sample for which the conditions from when the sample is collected from the subject to when the expression levels of a plurality of small RNAs in the sample are measured satisfy predetermined conditions, as a sample to be evaluated.

The disease determination unit 170 executes the disease determination step 116 described above. That is, the disease determination unit 170 determines the presence or absence of a disease by inputting small RNA data indicating the expression levels of multiple small RNAs in the selected specimen into the disease-trained model generated by the generation method 10 described above.

In this embodiment, the property determination system 20 includes a measurement device 21 and a disease determination device 30, but the property determination system 20 may be configured with a single device.

The disease assessment device 30 may also be composed of multiple devices. For example, the disease assessment device 30 may be composed of multiple (e.g., two) devices that share the tasks of performing the above-mentioned selection process 114 and the disease assessment process 116.

<Effects of this embodiment>
The method for generating a disease-trained model according to this embodiment is, as described above, a method for selecting a sample from which multiple small RNAs are derived when generating a disease-trained model for determining a disease in a subject based on the expression levels of multiple small RNAs, in which a sample that satisfies predetermined conditions (hereinafter referred to as selection conditions) from the time the sample is collected from the subject to the time the expression levels of multiple small RNAs in the sample are measured is selected as a standard sample, and a disease-trained model is generated using the expression levels of multiple small RNAs in the standard sample.
Therefore, it is possible to generate a disease-trained model with higher accuracy of judgment compared to a case where a disease-trained model is generated using the expression levels of multiple small RNAs in a sample without selecting the standard sample.

In addition, in this embodiment, the selection condition is the collection processing time from when a bodily fluid as a sample is collected from a subject to when centrifugation is completed, and in this generation method, a sample with an appropriate collection processing time is selected as a standard sample, and a disease-learned model is generated using the expression levels of multiple small RNAs in the standard sample.
Therefore, regardless of the appropriateness of the collection processing time, a disease-trained model with higher determination accuracy can be generated compared to a case in which a sample is selected as a standard sample and a disease-trained model is generated using the expression levels of multiple small RNAs in the standard sample.

Furthermore, in this embodiment, the small RNA data is input to a collection processing time trained model that has been machine-learned to learn the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and collection processing time data showing the collection processing time of the body fluids collected from the multiple subjects, thereby testing the appropriateness of the collection processing time and selecting samples with appropriate processing times as standard samples. In this way, a collection processing time trained model is used that uses not only measurement data from healthy subjects but also measurement data from subjects suffering from a disease as training data, so that the appropriateness of the collection processing time can be tested with high accuracy and samples with inappropriate collection processing times can be prevented from being selected as standard samples.

In addition, in this embodiment, the selection conditions are the storage temperature of the bodily fluid as a sample collected from the subject and the storage time at that storage temperature, and in this generation method, a sample with an appropriate storage temperature and storage time is selected as a standard sample, and a disease-learned model is generated using the expression levels of multiple small RNAs in the standard sample.
Therefore, regardless of the appropriateness of the storage temperature and storage time, it is possible to generate a disease-trained model with higher accuracy than when a sample is selected as a standard sample and a disease-trained model is generated using the expression levels of multiple small RNAs in the standard sample.

Furthermore, in this embodiment, the small RNA data is input into a storage condition learned model that has been machine-learned to learn the correlation between measurement data indicating the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and storage condition data indicating the storage temperature of the body fluids collected from the multiple subjects and the storage time at that storage temperature, thereby inspecting the appropriateness of the storage temperature and storage time at that storage temperature, and selecting samples with appropriate storage temperatures and storage times as standard samples.
In this way, a storage condition trained model is used that uses not only measurement data from samples from healthy subjects but also measurement data from samples from diseased subjects as training data, so that the appropriateness of the storage temperature and storage time can be inspected with high accuracy, and samples with inappropriate storage temperatures and times can be prevented from being selected as standard samples.

In addition, in this embodiment, the selection condition is the degree of hemolysis of the blood sample collected from the subject, and in this generation method, a sample with an appropriate degree of hemolysis is selected as a standard sample, and a disease-trained model is generated using the expression levels of multiple small RNAs in the standard sample. Therefore, a disease-trained model with higher determination accuracy can be generated compared to a case in which a sample is selected as a standard sample regardless of the appropriateness of the degree of hemolysis, and a disease-trained model is generated using the expression levels of multiple small RNAs in the standard sample.

Furthermore, in this embodiment, the degree of hemolysis is tested by inputting small RNA data into a hemolysis degree trained model that has been machine-learned to learn the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in blood collected from multiple subjects and hemolysis degree data showing the degree of hemolysis of the blood collected from the multiple subjects, and a sample with an appropriate degree of hemolysis is selected as a standard sample. In this way, the degree of hemolysis of blood is tested by inputting small RNA data into a hemolysis degree trained model that has been machine-learned to learn the correlation between measurement data and hemolysis degree data, so that the degree of hemolysis of blood can be tested with high accuracy compared to, for example, testing by visual inspection or measuring Hb concentration using a measurement kit, and samples with an inappropriate degree of hemolysis can be prevented from being selected as standard samples.

<Example>
Next, examples will be described. Note that the examples are merely examples of the technology of the present disclosure, and the technology of the present disclosure is not limited to the contents of the examples.

Example 1
In Example 1, the collection step 11 and the storage step 13 were appropriately performed for 64 lung cancer patients to obtain serum free from hemolysis. That is, for the 64 lung cancer patients, serum free from hemolysis was obtained as a standard specimen, which was collected and processed within 2 hours and stored at -80°C under appropriate storage conditions (storage time and temperature). The expression levels of multiple small RNAs were measured for the serum using NGS, and small RNA data showing the measurement results were obtained.
In Example 1, the collection step 11 and the storage step 13 were appropriately performed for 133 healthy subjects to obtain serum free of hemolysis. That is, serum free of hemolysis was obtained as a standard specimen from 133 healthy subjects, which was collected and processed within 2 hours and stored at -80°C under appropriate storage conditions (storage time and storage temperature). The expression levels of multiple small RNAs in the serum were measured using NGS, and small RNA data showing the measurement results was obtained.
Then, these small RNA data were used as training data to generate a disease-trained model 1 that determines the disease of a subject.

Comparative Example 1
In Comparative Example 1, serum was obtained from 64 lung cancer patients in the same manner as in Example 1. That is, serum was obtained from 64 lung cancer patients, with the collection processing time being within 2 hours, and the serum was stored under appropriate storage conditions (storage time and storage temperature) at -80°C, and free of hemolysis. The serum was measured for the expression levels of multiple small RNAs using NGS, and small RNA data showing the measurement results was obtained.
In Comparative Example 1, serum was obtained from 109 healthy subjects in the same manner as in Example 1. That is, serum was obtained from 109 healthy subjects, with the collection and processing time being within 2 hours, and the serum was stored at -80°C under appropriate storage conditions (storage time and storage temperature), and free of hemolysis.
Furthermore, in Comparative Example 1, serum free from hemolysis was obtained from 24 healthy subjects, which had been collected and processed for 24 hours (more than 2 hours). The serum was stored at -80°C, which is an appropriate storage condition (storage time and storage temperature). The serum was measured for the expression levels of multiple small RNAs using NGS, and small RNA data showing the measurement results was obtained.
Then, these small RNA data were used as training data to generate a disease-trained model 2 that determines the disease of the subject.

Comparative Example 2
In Comparative Example 2, serum was obtained from 64 lung cancer patients in the same manner as in Example 1. That is, serum was obtained from 64 lung cancer patients, with the collection and processing time being within 2 hours, and the serum was stored under appropriate storage conditions (storage time and storage temperature) at -80°C, and free of hemolysis. Expression levels of multiple small RNAs were measured for the serum using NGS, and small RNA data showing the measurement results were obtained.
In Comparative Example 2, serum was obtained from 109 healthy subjects in the same manner as in Example 1. That is, serum was obtained from 109 healthy subjects, with the collection and processing time being within 2 hours, and the serum was stored at -80°C under appropriate storage conditions (storage time and storage temperature), and free of hemolysis.
Furthermore, in Comparative Example 2, serum free from hemolysis was obtained from 24 healthy subjects that had been stored under inappropriate storage conditions (storage temperature of 4° C. for 24 hours). The serum was collected and processed within 2 hours. The serum was measured for the expression levels of multiple small RNAs using NGS, and small RNA data showing the measurement results was obtained.
Then, these small RNA data were used as training data to generate a disease-trained model 3 that determines the disease of the subject.

Comparative Example 3
In Comparative Example 3, serum was obtained from 64 lung cancer patients in the same manner as in Example 1. That is, serum was obtained from 64 lung cancer patients, with the collection processing time being within 2 hours, and the serum was stored under appropriate storage conditions (storage time and storage temperature) at -80°C, and free of hemolysis. Expression levels of multiple small RNAs were measured for the serum using NGS, and small RNA data showing the measurement results were obtained.
In Comparative Example 3, serum was obtained from 109 healthy subjects in the same manner as in Example 1. That is, serum was obtained from 109 healthy subjects, with the collection and processing time being within 2 hours, and the serum was stored at -80°C under appropriate storage conditions (storage time and storage temperature), and free of hemolysis.
Furthermore, in Comparative Example 3, hemolyzed serum was obtained from 24 healthy subjects. The serum was collected and processed within 2 hours and stored at -80°C under appropriate storage conditions (storage time and temperature). The serum was measured for expression levels of multiple small RNAs using NGS, and small RNA data showing the measurement results was obtained.
Then, these small RNA data were used as training data to generate a disease-trained model 4 that determines the disease of the subject.

<Example of learning model>
In generating a disease trained model, various linear and nonlinear algorithms known as machine learning algorithms, or a combination of multiple algorithms, can be used. For example, the following algorithms can be used:

Random forest
Dropout Additive Regression Trees
Gradient boosted trees
Extreme gradient boosted trees
Light-gradient boosted machine
Neural networks
Regularized regression
Elastic-net
K-Nearest neighbors
Support vector machine
Generalized additive model

<Judgment results for disease-trained models 1 to 4>
In the disease trained model 1 of Example 1, for example, the trained model “Generalized additive
When the "disease trained model" was used, the AUC was 0.957, which was better than the disease trained model 2 of Comparative Example 1 (0.819), the disease trained model 3 of Comparative Example 2 (0.896), and the disease trained model 4 of Comparative Example 3 (0.900) (see FIG. 5). Similarly, as shown in FIG. 5, even when other types of learning models were used, the disease trained model 1 of Example 1 was better than the disease trained model 2 of Comparative Example 1, the disease trained model 3 of Comparative Example 2, and the disease trained model 4 of Comparative Example 3.

The present invention is not limited to the above-described embodiment, and various modifications, changes, and improvements are possible without departing from the spirit of the invention. The above-described modifications may be combined as appropriate.

<Additional Notes>
(Aspect 1)
A method for generating a property determination model for determining the property of a subject based on expression levels of a plurality of small RNAs, the method comprising the steps of: selecting a specimen from which the plurality of small RNAs are derived;
A sample that satisfies a predetermined condition from the time of collection of the sample from the subject to the time of measuring the expression levels of a plurality of small RNAs in the sample is selected as a standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property judgment model.
(Aspect 2)
The standard sample is a sample in which at least one of the following conditions is satisfied in advance: collection conditions for collecting a sample from a subject, storage conditions for storing the collected sample, and measurement conditions for measuring the expression levels of a plurality of small RNAs in the subject.
A method for generating a property determination model according to aspect 1.
(Aspect 3)
the condition is a processing time from collection of the body fluid as the sample from the subject to completion of a centrifugation operation,
A sample having an appropriate treatment time is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to aspect 2.
(Aspect 4)
The suitability of the processing time is examined by inputting small RNA data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of the subjects into a processing time trained model that has been machine-learned to determine the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and processing time data showing the processing time of the body fluids collected from the multiple subjects;
A sample having an appropriate treatment time is selected as the standard sample;
generating a trait determination model for determining a disease as the trait of a subject using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to aspect 3.
(Aspect 5)
the condition is a storage temperature of the body fluid as the sample collected from the subject and a storage time at the storage temperature;
A sample having the appropriate storage temperature and storage time is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to any one of aspects 2 to 4.
(Aspect 6)
a storage condition trained model that has been machine-learned to learn a correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of multiple subjects, including subjects suffering from a disease, and storage condition data showing the storage temperature of the body fluids collected from the multiple subjects and the storage time at that storage temperature, and inputs small RNA data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of the subjects into the model, thereby examining the suitability of the storage temperature and the storage time at that storage temperature;
A sample having the appropriate storage temperature and storage time is selected as the standard sample;
generating a trait determination model for determining a disease as the trait of a subject using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to aspect 5.
(Aspect 7)
the condition is a degree of hemolysis of the blood sample collected from the subject,
A sample having an appropriate degree of hemolysis is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to any one of aspects 2 to 6.
(Aspect 8)
A hemolysis degree trained model is configured to machine-learn a correlation between measurement data showing the results of measuring the expression levels of a plurality of small RNAs in blood collected from a plurality of subjects and hemolysis degree data showing the hemolysis degree of the blood collected from the plurality of subjects, by inputting small RNA data showing the results of measuring the expression levels of a plurality of small RNAs in the body fluid of the subject, and examining the degree of hemolysis;
A sample having an appropriate degree of hemolysis is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property determination model according to aspect 7.
(Aspect 9)
Generate a property determination model for determining a disease as the property of the subject;
A method for generating a property determination model according to any one of aspects 1 to 8.
(Aspect 10)
A property determination model for determining a property of a subject, the property determination model being generated based on expression levels of a plurality of small RNAs, and a specimen from which the plurality of small RNAs are derived being selected during the generation,
A sample that satisfies predetermined conditions from collection of the sample from the subject to measurement of the expression levels of multiple small RNAs in the sample is selected as a standard sample;
A property determination model generated using the expression levels of the plurality of small RNAs in the standard specimen.
(Aspect 11)
A method for determining a property, comprising inputting small RNA data showing the results of measuring expression levels of a plurality of small RNAs in a specimen of a subject into the property determination model according to aspect 10, thereby determining the property.
(Aspect 12)
The small RNA data is
A property determination method according to aspect 11, wherein the data obtained from the sample satisfies the predetermined conditions from when the sample is collected from the subject until when the expression levels of the plurality of small RNAs in the sample are measured.
(Aspect 13)
A property determination model according to aspect 10,
The property determining device determines the property by inputting small RNA data showing the results of measuring the expression levels of a plurality of small RNAs in the body fluid of the subject into the property determining model.

The disclosure of Japanese Patent Application No. 2023-182776, filed on October 24, 2023, is incorporated herein by reference in its entirety. All documents, patent applications, and technical standards described herein are incorporated herein by reference to the same extent as if each individual document, patent application, and technical standard was specifically and individually indicated to be incorporated by reference.

Claims (13)

A method for generating a property determination model for determining the property of a subject based on expression levels of a plurality of small RNAs, the method comprising the steps of: selecting a specimen from which the plurality of small RNAs are derived;
A sample that satisfies a predetermined condition from the time of collection of the sample from the subject to the time of measuring the expression levels of a plurality of small RNAs in the sample is selected as a standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
A method for generating a property judgment model. The standard sample is a sample in which at least one of the following conditions is satisfied in advance: collection conditions for collecting a sample from a subject, storage conditions for storing the collected sample, and measurement conditions for measuring the expression levels of a plurality of small RNAs in the subject.
A method for generating a property determination model according to claim 1 . the condition is a processing time from collection of the body fluid as the sample from the subject to completion of a centrifugation operation,
A sample having an appropriate treatment time is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 2 . The suitability of the processing time is examined by inputting small RNA data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of the subjects into a processing time trained model that has been machine-learned to determine the correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in body fluids collected from multiple subjects, including subjects suffering from a disease, and processing time data showing the processing time of the body fluids collected from the multiple subjects;
A sample having an appropriate treatment time is selected as the standard sample;
generating a trait determination model for determining a disease as the trait of a subject using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 3 . the condition is a storage temperature of the body fluid as the sample collected from the subject and a storage time at the storage temperature;
A sample having the appropriate storage temperature and storage time is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 2 . a storage condition trained model that has been machine-learned to learn a correlation between measurement data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of multiple subjects, including subjects suffering from a disease, and storage condition data showing the storage temperature of the body fluids collected from the multiple subjects and the storage time at that storage temperature, and inputs small RNA data showing the results of measuring the expression levels of multiple small RNAs in the body fluids of the subjects into the model, thereby examining the suitability of the storage temperature and the storage time at that storage temperature;
A sample having the appropriate storage temperature and storage time is selected as the standard sample;
generating a trait determination model for determining a disease as the trait of a subject using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 5 . the condition is a degree of hemolysis of the blood sample collected from the subject,
A sample having an appropriate degree of hemolysis is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 2 . A hemolysis degree trained model is configured to machine-learn a correlation between measurement data showing the results of measuring the expression levels of a plurality of small RNAs in blood collected from a plurality of subjects and hemolysis degree data showing the hemolysis degree of the blood collected from the plurality of subjects, by inputting small RNA data showing the results of measuring the expression levels of a plurality of small RNAs in the body fluid of the subject, and examining the degree of hemolysis;
A sample having an appropriate degree of hemolysis is selected as the standard sample;
generating the property determination model using the expression levels of the plurality of small RNAs in the standard specimen;
The method for generating a property determination model according to claim 7. Generate a property determination model for determining a disease as the property of the subject;
A method for generating a property determination model according to claim 1 . A property determination model for determining a property of a subject, the property determination model being generated based on expression levels of a plurality of small RNAs, and a specimen from which the plurality of small RNAs are derived being selected during the generation,
A sample that satisfies predetermined conditions from collection of the sample from the subject to measurement of the expression levels of a plurality of small RNAs in the sample is selected as a standard sample;
A property determination model generated using the expression levels of the plurality of small RNAs in the standard specimen. A property determination method comprising: inputting small RNA data showing the results of measuring expression levels of a plurality of small RNAs in a specimen of a subject into the property determination model according to claim 10, thereby determining the property. The small RNA data is
The method for determining a property according to claim 11 , wherein the data obtained from the sample satisfies the predetermined conditions from when the sample is collected from the subject to when the expression levels of multiple small RNAs in the sample are measured. The property determination model according to claim 10,
The property determining device determines the property by inputting small RNA data showing the results of measuring the expression levels of a plurality of small RNAs in the body fluid of the subject into the property determining model.